Evaluation Of Topographic Wind Energy Resources And Correction Of Wind Speed Forecast In A Desert Steppe

As a renewable clean energy,wind energy is more and more to the attention of the people,but because the wind is affected by wind,and the wind is random,intermittent and uncertain characteristics,seriously interfere with the safe and stable operation of power grid,which result in the output power of wind power fluctuation is very big,the wind power grid to bring very great challenge.In this paper,the key scientific problems in wind energy development:wind resource assessment of complex terrain and short-term wind speed forecast are studied.There are many similarities between wind energy resource assessment and short-term wind speed prediction theory and research methods in complex terrain.Wind energy development is not only a practical engineering problem that must be faced by wind energy resource development,but also a challenge subject closely connected with basic research and application of atmospheric science.In order to meet the needs of wind energy assessment of large-scale wind farms under complex terrain conditions,the main research work of this paper is determined as follows:to establish the coupling model system of"WRF+ARTIFICIAL intelligence",and to carry out refined wind energy resource assessment and short-term wind speed forecast research,which is especially suitable for large-scale wind farms under complex terrain conditions.The main contents are as follows:The complex and changeable terrain has a great influence on the distribution of wind energy resources,which leads to the uncertainty in the evaluation of wind energy resources.Based on the meteorological tower data of different desert steppe terrain backgrounds in northern China from 2018 to 2020,the influences of Kernel,Weibull and Rayleigh wind speed distributions on the estimation of average wind energy density were compared.Then,the Weibull model is used to estimate wind energy resources based on three key parameters:scale factor(c),shape factor(k)and surface roughness(z₀).The results show that Weibull distribution is the most suitable wind speed distribution for this terrain.The scale factor(c)in Weibull distribution model increases with the increase of height,showing an obvious power function form.The relation between shape factor(k)and height has two different forms:reciprocal of quadratic function and reciprocal of logarithmic function respectively.Roughness(z₀)varies with the wilting,growing,and luxuriant stages and can be expressed as the median of estimates for each period.The maximum surface roughness length is 0.15m and the minimum is 0.12m in the whole cycle.The power law model and logarithmic model are used to estimate the average power density values of six specific heights,and the differences are large in autumn and winter,but small in spring and summer.The gradient of average power density value increasing with height is maximum in autumn and winter,and minimum in spring and summer.The results show that the dynamic changes of three key parameters(c,k and z₀)should be accurately considered when estimating wind energy resources in different desert steppe terrains.The accurate prediction of short-term wind speed can obtain more accurate distribution of wind energy resources,which can promote the large-scale development of wind power generation and reduce the influence of the uncertainty of wind power generation on the operation of power grid.By August 2021 to February 2022 in the five provinces of the south of mesoscale weather forecast model(WRF)10 meters 24 h wind speed forecasting the forecast on the limitation of data in the region and the same time of the observation data of 410 weather stations comparison found that wind speed forecast accuracy improvement,this paper use four machine learning method(deep belief network(DBN),Multi-layer sensors(MLP),support vector regression(SVR)and random forest(RF)were used to revise the 10 m wind speed forecast by WRF model.The 10 m wind speed forecast and actual data in February 2022 were used as the training set,and the 10 m wind speed forecast and actual data were used as the test set from 00 o'clock on March 1,2022 to 15 o'clock on March 7,2022.The results show that the stochastic forest model optimized by Bayesian algorithm is the most suitable integrated model for wind speed prediction in the five provinces of south China.The prediction accuracy is up to 89%,and the root mean square error of prediction is as low as 0.43m/s.